Speaker
Description
Stripper foil degradation at the CERN Low Energy Ion Ring (LEIR) poses a significant challenge for beam operations. As the heavy ion beam passes through the stripper foil at the end of the injecting linac, the foil degrades over time, altering the beam energy distribution and reducing the achievable accumulated intensity in the ring. Addressing this operational limitation using traditional control approaches is challenging due to the complex, multi-dimensional nature of the multi-turn injection process. This paper presents a reinforcement learning-based controller to compensate for foil degradation and maintain ring performance. The controller observes longitudinal Schottky spectra encodings and time-of-flight measurements from the linac to adjust the ramping and debunching cavity phases, and electron cooler gun and orbit bump in real-time. We demonstrate that pre-training the agent in a data-driven surrogate model significantly improves both controller performance and sample efficiency during deployment.
| In which format do you inted to submit your paper? | LaTeX |
|---|